Writing
Reporting
3 min read
By James Sheriff
One report from five tools: pulling your business data together
Sales are in one tool, costs in another, enquiries in a third. Every month someone copies the numbers into a spreadsheet so the business can see how it's doing. That job can go.
Contents
The monthly spreadsheet
Most small businesses have one. Someone exports figures from the accounts software, the shop, the CRM and the website analytics, pastes them into a spreadsheet, fixes the formulas that broke since last month, and sends it round.
It takes hours. It’s out of date the moment it’s sent. And if the person who builds it is off, nobody quite knows how.
Start with the questions, not the data
The most common mistake is starting with “let’s get all our data in one place”. You end up with a lot of data and no clearer answers.
Start instead with the handful of questions the business actually asks. For example:
- How many enquiries did we get this month, and from where?
- How many turned into sales?
- Which products or services make the most margin?
- Are we busier or quieter than this time last year?
Each question tells you which tools you need data from, and which numbers in those tools matter. Everything else can wait.
How it works
There are three parts.
1. Collect
Most business software lets you get your data out automatically, usually through an API. A small program collects the figures from each tool on a schedule, every night for example, and stores them in one database.
Where a tool has no way to get data out automatically, there’s usually still a regular export that can be picked up.
2. Define each number once
This is the part that makes reports trustworthy. “A sale” might mean an order placed in the shop, an invoice raised in the accounts, or a deal marked as won in the CRM. They won’t match.
Decide once what each number means and where it comes from, write it down, and build it into the report. Then everyone is arguing about the business, not about whose spreadsheet is right.
3. Show it
A dashboard tool turns the database into charts and tables that update themselves. I’ve used Metabase for this: it has an open-source edition that can run on your own server, so the data stays under your control.
Reports can be a page you check whenever you like, or an email that arrives every Monday morning.
Where AI fits, and where it doesn’t
The core of this is plain data work: collecting, matching, adding up. It doesn’t need AI, and numbers you rely on shouldn’t be guessed by a language model.
AI can help around the edges: tidying up messy data, such as matching “J Smith Ltd” to “John Smith Limited” across two systems, or writing a short summary of what changed this month alongside the figures. The figures themselves should come straight from the source.
What you end up with
- The same numbers for everyone, defined the same way.
- Reports that update themselves instead of being rebuilt.
- The hours that went into the monthly spreadsheet, back.
Is it worth it?
If someone spends a day or more a month building reports by hand, probably. If your business runs from one tool that already reports well, probably not.
Which jobs are worth automating has a simple test, and what AI automation costs covers the prices.